|  | Anxiety vs. use/no useN= 15 (Figure2) | Anxiety vs. CUD/no CUD (or no use)N= 13 (Figure3) | Anxiety + depression vs. use/no useN= 5 (Figure4) |
---|---|---|---|---|
Overall mean weighted OR | OR (95% CI) p two-tailed | 1.24 (1.06-1.45) .006* | 1.68 (1.23-2.31) .001* | 1.68 (1.17-2.40) .004* |
Heterogeneity statistics | Q(df) p two-tailed I 2 | 30 (14) .009* 53% | 55 (12) <.0001* 78% | 8 (4) .091 50% |
One-study removed analysis (Additional file 1 : Figure S1) | Which studies, removed one at a time from the analysis, remove the significance of the overall mean weighted OR? | None | None | Cheung et al., 2010 [36] |
Cumulative analysis (1993 to 2013) (Additional file 1 : Figure S2) | Which studies, added to all previous studies one at a time, remove the significance of the overall mean weighted OR? | Degenhardt et al., 2001 [39] | NPMS, 2007 [10] | |
Rosenthal’s Fail-safe N for p > .05 | N studies needed to remove the significance of the overall mean weighted OR (N studies missing for every study in meta-analysis needed to remove the significance of the overall mean weighted OR) | N=49 (49/15 = 3) | N=105 (105/13 = 8) | N=17 (17/5 = 3) |
Duval and Tweedie’s Trim-and-Fill analysis | Funnel plot symmetrical? N studies missing on either side of the overall mean weighted OR | YES | YES | YES |
N=0 | N=0 | N=0 | ||
Begg and Mazumdar Rank Order Correlation | Ï„ | .13 | -.19 | -.30 |
p two-tailed | .488 | .360 | .462 | |
Egger’s regression | intercept | .25 | 1.73 | −4.05 |
 | p two-tailed | .748 | .005* | .415 |